xAI currently operates Colossus, one of the world's largest AI supercomputers with 100,000+ Nvidia GPUs in Memphis, scaling toward 1 million GPUs. The company is also developing additional compute facilities, like Colossus 2. This massive infrastructure buildout, representing tens of billions in investment, could position xAI to sell excess compute capacity similar to how AWS emerged from Amazon's internal needs.

Major AI companies (OpenAI, Anthropic) rely on cloud providers rather than owning infrastructure. Selling excess compute could offset massive capital expenditures.

This market predicts whether xAI will transition from pure AI development to also becoming an infrastructure provider.

Resolution Criteria

Resolves YES if by 11:59 PM PT on December 31, 2026:

xAI offers AI compute services (training or inference) to external customers through ANY of the following:

1. Public cloud service offering:

   • Publicly accessible platform where customers can purchase/rent compute

   • Website or portal allowing sign-ups for compute access

   • Published pricing for compute hours/tokens

2. Enterprise/Partnership deals:

   • Announced deals to provide compute to named companies

   • SEC filings mentioning compute service revenue

   • Earnings reports showing infrastructure-as-a-service revenue

3. Official confirmation via:

   • xAI website, blog, or press releases

   • Elon Musk on X/Twitter explicitly confirming external compute sales

   • Customer testimonials or case studies on xAI channels

Resolves NO if:

• No external compute offerings by deadline

• Only internal use for xAI/Grok development

• Only partnerships where xAI receives compute (not provides)

• Vague statements about "considering" or "exploring" without actual availability

Important Clarifications:

✅ DOES count:

• Any external customer (not xAI/Tesla/Musk companies) purchasing compute

• Compute for training, fine-tuning, or inference

• Beta programs with paying customers

• Revenue-sharing deals where partners get compute in exchange for payment

• Reselling compute from their infrastructure (even if not all hardware owned)

• "Grok API" services that charge for compute/tokens (not just model access)

❌ Does NOT count:

• Free research partnerships or grants

• Internal use by other Elon Musk companies (Tesla, X, SpaceX, etc.)

• Only licensing Grok model without compute infrastructure

• Collaboration where both parties contribute resources

• Selling hardware (must be compute services)

• Traditional software/API licensing without compute component

Special Cases:

Tesla Exception:

• If xAI formally merges with Tesla before deadline, compute sold under "Tesla" branding DOES count if the infrastructure is clearly the Memphis Colossus facility, or other explicitly xAI AI facilities (like Colossus 2)

API Distinction:

• Simple Grok API access (like ChatGPT API) does NOT count

• "Bring your own model" training services DO count

• Must be infrastructure/compute, not just model-as-a-service

Proof Requirements:

• Need evidence of actual availability, not just announcements of future plans

• At least one confirmed external customer or public sign-up availability

• "General availability in 2027" announcements do NOT qualify

• Update 2026-04-24 (PST) (AI summary of creator comment): A potential deal involving SpaceXAI paying Cursor $10B (or acquiring them) does NOT count toward YES resolution, as it falls under the "Collaboration where both parties contribute resources" exclusion. The market remains unresolved as NO for now.

• Update 2026-05-06 (PST) (AI summary of creator comment): The Anthropic/SpaceX compute deal is being evaluated as a potential YES resolution. The creator is waiting a few days to determine whether it is a straightforward cash-for-compute arrangement (which would resolve YES) or more of a resource-sharing partnership (which would not count, similar to the Cursor exclusion).